The cloud offers a completely new cost model for IT based on elasticity and a pay-as-you-go approach. Most companies...
have to rethink their project planning to get the most out of cloud, but many don't realize how easy it is to get too much cloud -- a challenge known as cloud sprawl.
And, unless you take steps to control it, cloud sprawl can derail your project plans.
Cloud sprawl is the proliferation of cloud resources that makes a deployment unnecessarily expensive. Companies often cite three reasons for cloud sprawl: a lack of central coordination over cloud resources, suboptimal use of multiple cloud providers and overprovisioning resources for an application. All of these, however, can be avoided with careful cloud resource management.
Central coordination, policy creation is key
Cloud computing is an expense, not a capital project. Many companies don't apply stringent purchase control on IT services like cloud computing, which means either IT groups or line departments can acquire new cloud services without much review.
A company's public cloud services form a virtual data center, which needs to be optimized for collective use. That doesn't mean central IT teams have to manage all public cloud use, but they should set basic guidelines to ensure users aren't being wasteful with cloud resources. This requires a cloud policy.
A cloud policy should define the smallest possible number of cloud models -- such as infrastructure as a service, platform as a service and software as a service, or even VMs and containers -- that can support the company's applications. Then, select prime and backup public cloud providers that can supply those models. Require both IT and line organizations using public cloud services to pick from the providers and models identified in the plan.
Multiple providers can increase cloud sprawl risks
Multiple providers potentially increase sprawl, and complicate cloud resource management, because of service redundancy. Many cloud provider plans require a minimum resource commitment, especially plans that involve dedicated or allocated hosting. If two organizations pick different providers, each will have to meet the minimums, and any excess resources may be unavailable to others.
This problem shows that there is more to preventing cloud sprawl than just setting a policy. Someone or some organization has to be responsible for monitoring all cloud commitments to ensure the policies are not violated. This team should also combine all cloud contracts with a single provider to get the best commercial terms possible. A central IT department is best for this, but companies without one can have someone that reports to a senior executive, such as the chief operating officer, handle contract coordination.
The task of cloud coordination is really a task of acquisition and allocation. When you need a new cloud service, instead of providing it through a new contract, assign cloud capacity from a suitable capacity pool created by prior contracts. If this pool is insufficient, then augment it by either extending current contracts or getting new ones -- but no application should own the cloud resources; they're owned by the pool.
Beware of overprovisioning
Even these steps won't completely eliminate cloud sprawl, as it's still possible to overcommit resources to an application and create waste. This is most likely to be a problem when line departments with no IT skills negotiate cloud hosting contracts or deal with a third-party reseller or integrator. For example, an application might run in a dedicated cloud server or have high-availability features when the actual application requirements don't justify those capabilities.
The problem with resource commitments in cloud services is that it's difficult to know what the right level is. Require an application trial based on a short-term commitment to the most basic cloud service available, or one that uses cloud resources within the company pool. If the trial shows issues in QoE, then additional cloud features are justified.
The final steps to avoid sprawl, and optimize cloud resource management, are ongoing reviews of billing and contract renewals. Public cloud users often add to cloud services without a formal review of their needs and the service costs. To avoid this, you'll need some cooperation from the CFO or accounts payable, as new charges are often hard to dig out of complex cloud invoices. Any new charge should be justified.
Cloud's pay-as-you-go model can result in too much going and too much paying. Companies should recognize that inefficient use and the risk of overpaying is the inevitable consequence of distributed cloud decisions. Tackle these challenges early on to have a better cloud resource management plan.
Editor's note: This article is part one in a two-part series on identifying and preventing cloud sprawl. You can read part two here.
Form a cloud capacity plan
Avoid the costly repercussions of overprovisioning
Control cloud sprawl's effect on performance